Wet Removal and Cloud Enhancement: The Microphysics of Cloud-Haze Interactions on Sub-Neptunes

Abstract: Aerosols are a near-ubiquitous feature of sub-Neptune atmospheres, yet their microphysical nature remains poorly understood. Both condensate clouds and photochemical hazes have been proposed to explain observations, but have largely been studied in isolation. Here we present a new bin-scheme microphysical model, adapted from CARMA, that couples cloud and haze formation through heterogeneous nucleation - the dominant mode of cloud formation in the Solar System - in which haze particles act as cloud condensation nuclei (CCN). Applying this model to KCl clouds on GJ 1214 b-like warm sub-Neptunes, we find that the microphysical contact angle $θ$ between cloud and haze particles governs distinct regimes of aerosol behavior: at moderate contact angles ($25^\circ \lesssim θ\lesssim 70<sup>\circ$), hazes are efficiently removed from the upper atmosphere through "wet removal" as they seed gravitationally-settling clouds; at small contact angles ($θ\lesssim 25<sup>\circ$), heterogeneous nucleation instead produces an enhanced population of mixed cloud-haze particles at high altitudes, dramatically increasing aerosol optical depth ("cloud enhancement"). These structural changes produce differences of up to four scale heights in transmission spectra, with strong effects at optical and near-infrared wavelengths relevant to JWST NIRISS/SOSS, while wavelengths beyond about 3 microns remain comparatively unaffected. We map these effects across orders of magnitude in metallicity, haze production rate, and vertical mixing strength, establishing their generality across sub-Neptune parameter space. Because heterogeneous nucleation is a universal phase-change process, this framework extends naturally to other exoplanet atmospheres and potentially any astrophysical environments where condensation onto foreign substrates may occur, including protoplanetary disks and stellar outflows.

• The paper develops a coupled CARMA model of KCl clouds, photochemical hazes, and haze-seeded mixed clouds, finding that contact angle creates wet-removal and cloud-enhancement regimes.
• The simulations show that strong cloud–haze interactions can increase upper-atmosphere cloud mass nearly twofold, boost particle numbers by about 25 times, or clear aerosols through rainout.
• Synthetic spectra vary by 2–5 scale heights below 2.5 μm, demonstrating that uncertain contact angles and vertical mixing can mimic changes in metallicity or atmospheric composition.

Motivation: clouds and hazes in isolation have failed

Sub-Neptune transmission spectra are routinely muted by aerosols, with GJ 1214 b as the canonical case: a decade of observations yielded an essentially featureless spectrum (Kreidberg et al., 2013), while JWST phase-curve work indicates a Bond albedo near 0.4 and a metallicity exceeding 1000× solar (Kempton et al., 2023). Existing microphysical models struggle to reconcile this with either aerosol type alone. Cloud-only models require $K_{zz} > 10^{10}\ \mathrm{cm^2\ s^{-1}}$ to loft KCl particles to observable altitudes—orders of magnitude above GCM-predicted values (Gao et al., 2018)—while haze-only models require production rates $>10^{-9}\ \mathrm{g\ cm^{-2}\ s^{-1}}$, far above photolysis-based estimates (Ohno et al., 2024), and tholin-like hazes are too absorbing to explain the high albedo. The paper's premise is that these two populations should not be modeled independently: across the Solar System, cloud formation proceeds overwhelmingly through heterogeneous nucleation on pre-existing seeds, so photochemical haze particles plausibly serve as cloud condensation nuclei (CCN) in sub-Neptune atmospheres.

A coupled bin-scheme microphysical framework

The authors extend CARMA into a unified framework tracking three aerosol populations: homogeneously nucleated "pure" KCl clouds, photochemical hazes (Titan-like tholins), and "mixed" KCl clouds nucleated heterogeneously on haze CCN. The simulations use 1D P–T profiles averaged from GCMs of GJ 1214 b with radiatively active hazes (Steinrueck et al., 28 Mar 2025), power-law $K_{zz}$ profiles normalized to preserve the mixing timescale across metallicities, and equilibrium KCl abundances from FastChem evaluated at the condensation curve intersection rather than assuming the abundance tracks elemental potassium.

The central new parameter is the microphysical contact angle $\theta$ between condensate and haze substrate, which enters classical heterogeneous nucleation theory through the shape factor; smaller $\theta$ implies stronger cloud–haze affinity and exponentially higher nucleation rates at given supersaturation. The Kelvin effect couples this to particle size: only haze particles large enough to overcome the curvature-induced vapor pressure increase can act as CCN, and decreasing $\theta$ lowers that threshold size. The grid spans metallicities of 1–3000× solar, haze production rates from $10^{-19}$ to $10^{-11}\ \mathrm{g\ cm^{-2}\ s^{-1}}$, contact angles from 0.3° to 179.7°, and $K_{zz}$ scaled by 0.1×, 1×, and 10×.

Two caveats are stated plainly by the authors: realistic contact angles for exoplanet condensate–haze pairs are essentially unconstrained (laboratory measurements for KCl on haze analogs span roughly 0°–77° (Yu et al., 2021)), and the CARMA runs do not self-consistently feed back aerosol radiative effects—the fixed P–T profiles assume a single representative haze distribution from the underlying GCMs.

Two regimes governed by contact angle

The central result is a bifurcation in aerosol behavior as $\theta$ decreases:

Regime	Contact angle	Mechanism	Net effect
Wet removal	~25° ≲ θ ≲ 70°	Only large hazes act as CCN; mixed clouds grow, settle, evaporate; homogeneous nucleation suppressed	Upper atmosphere cleared of aerosols
Cloud enhancement	θ ≲ 25°	Small, abundant high-altitude hazes activated as CCN	Large mixed-cloud population aloft; enhanced opacity

In the wet removal regime—analogous to terrestrial nucleation scavenging—the largest haze particles are preferentially rained out, total cloud mass drops (by 25% between the grid endpoints in the fiducial case), and the atmosphere becomes clearer. In the cloud enhancement regime, activation of submicron haze seeds produces mixed clouds whose upper-atmosphere mass approaches twice the no-interaction value, with number densities increasing by factors of ~25 at high haze production rates. Notably, the response is non-monotonic in $>10^{-9}\ \mathrm{g\ cm^{-2}\ s^{-1}}$0: cloud mass passes through a minimum at moderate contact angles before rising sharply.

Metallicity modulates both regimes. At 1× solar, where homogeneous KCl nucleation is negligible, strong interactions still produce substantial high-altitude clouds—the cloud-mass mixing ratio at 0.1 mbar increases by more than two orders of magnitude between weak and strong interaction cases—whereas at 3000× solar the change is only ~2×. Haze production rate controls whether the system is seed-limited or seed-abundant: at low rates ($>10^{-9}\ \mathrm{g\ cm^{-2}\ s^{-1}}$1–$>10^{-9}\ \mathrm{g\ cm^{-2}\ s^{-1}}$2), interactions enlarge average particle sizes (e.g., from 0.2 to ~0.8 μm at 0.1 mbar) while reducing number density; at intermediate rates ($>10^{-9}\ \mathrm{g\ cm^{-2}\ s^{-1}}$3–$>10^{-9}\ \mathrm{g\ cm^{-2}\ s^{-1}}$4) the effect on observables is maximal; at the highest rates haze opacity dominates regardless of $>10^{-9}\ \mathrm{g\ cm^{-2}\ s^{-1}}$5.

Transmission spectral consequences

Synthetic spectra computed with PICASO 3.0 using full Mie scattering on the predicted size distributions show that cloud–haze interactions matter almost exclusively shortward of ~2.5 μm; wavelengths beyond ~3 μm are largely insensitive because upper-atmosphere particles rarely grow large enough. For low-to-moderate metallicities (1–300× solar) and intermediate haze production rates, varying $>10^{-9}\ \mathrm{g\ cm^{-2}\ s^{-1}}$6 alone spans 2–5 scale heights in transit depth, including 2–4 scale-height changes in the amplitude of the 1.4 μm water feature. This is a direct challenge to population-level studies that infer metallicity or aerosol trends from short-wavelength feature amplitudes: the paper demonstrates degeneracy between bulk atmospheric properties and unconstrained cloud–haze microphysics even when temperature, composition, and mixing are held fixed. At very high metallicity under fiducial mixing, spectra become nearly independent of contact angle—but this insensitivity is not robust, as the mixing experiments show.

The fractional-opacity analysis also shows that the dominant aerosol contributor is regime-dependent: with strong interactions ($>10^{-9}\ \mathrm{g\ cm^{-2}\ s^{-1}}$7), mixed clouds dominate the optical opacity from $>10^{-9}\ \mathrm{g\ cm^{-2}\ s^{-1}}$8 to $>10^{-9}\ \mathrm{g\ cm^{-2}\ s^{-1}}$9 in haze production rate, whereas with negligible interactions pure KCl clouds dominate below $K_{zz}$0. Contact angle can therefore qualitatively change which species one infers from an observed spectrum.

Sensitivity to vertical mixing

Factor-of-ten variations in $K_{zz}$1 propagate strongly through all results. A 10× increase raises upper-atmosphere cloud mass by more than three orders of magnitude at 3000× solar, and it restores contact-angle sensitivity to transmission spectra at high metallicity where the fiducial runs showed none. It also shifts the boundary between cloud-dominated and haze-dominated opacity by up to four orders of magnitude in haze production rate (from $K_{zz}$2 to $K_{zz}$3 between 0.1× and 10× $K_{zz}$4). Given that GCM-inferred $K_{zz}$5 carries order-of-magnitude uncertainties, this sensitivity means quantitative predictions from any single mixing prescription should be treated as illustrative rather than definitive.

Limitations and open questions

Several assumptions bound the results. The contact angle is treated as a free parameter spanning the full plausible range because laboratory constraints are sparse; the authors argue Solar System evidence (near-zero angles for water on common terrestrial CCN, small angles for methane and ethane on Titan tholins) suggests low values may be common, but dedicated measurements of condensate–haze surface energies for warm sub-Neptune compositions remain necessary. All particles are assumed spherical with tholin optical properties; fractal aggregates could lower effective contact angles via roughness and potentially enable condensation at subsaturation through concave ("inverse Kelvin") geometries, but this is unmodeled. The 1D framework omits horizontal advection, which 2D hot-Jupiter work shows can substantially increase upper-atmosphere cloud mass. Aerosol radiative feedback is inherited from a single representative GCM haze distribution rather than computed self-consistently. Finally, whether cloud–haze interactions produce a unique observational signature—beyond altering feature amplitudes—is left open, though heterogeneous particle optics and polarimetry are identified as candidate discriminators.

Conclusion

This paper establishes heterogeneous nucleation of condensates on photochemical haze as a first-order control on sub-Neptune aerosol distributions, organized by the microphysical contact angle into wet-removal and cloud-enhancement regimes. The resulting spectral variations—up to four scale heights at optical and near-infrared wavelengths, driven by microphysics alone—imply that short-wavelength transmission features cannot be straightforwardly interpreted as diagnostics of bulk atmospheric properties without accounting for cloud–haze coupling. The framework extends naturally to other condensate–substrate systems, from silicate clouds on hot Jupiters to methane clouds on Uranus and Neptune, wherever falling particles can serve as CCN.